Skip to content
MNMohammed Nisaruddin

About

Nine years of building software that has to work on real data.

I am Mohammed Nisaruddin, a Senior AI & Data Engineer based in Hyderabad, India. I design and build Generative AI applications, data pipelines and the Python backends that hold them together.

Download Resume Data engineering resumeGet in touch

Where it started

I began my career close to data: quality-checking training datasets for machine-learning models and using SQL to find fraud patterns across millions of transactions. It was not glamorous work, but it taught me something that has shaped everything since — systems are only as good as the data moving through them.

From there I moved into software engineering proper, building backend systems in Python and then data-heavy platforms: analytics pipelines, simulation data preparation, real-time ingestion from many sources.

Growing into AI engineering

When large language models became practical for real workloads, I was already working where they are most useful: large volumes of unstructured documents and the pipelines around them. Moving into Generative AI was less a career change than a natural extension — the same engineering discipline, applied to a new kind of component.

Since 2023 most of my work has been in the insurance and benefits domain, where I own the architecture and delivery of a document intelligence platform. It reads policy documents, identifies and classifies benefits, and produces structured, retrieval-ready data — work that used to take hours of manual reading per policy.

Python, backends and data platforms

Python is my working language. I build FastAPI services, integrate them with enterprise systems, and put authentication, queues and deployment around them. Alongside the AI work I contribute to a large banking data migration platform built on AWS Glue and PySpark, where correctness matters more than anything else.

That combination — AI engineering, backend engineering and data engineering — is what lets me take a problem from a conversation to a deployed system without handing it between three teams.

What I am focused on now

My current focus is agentic AI and the evaluation layer that keeps it trustworthy. In client work I have built MCP servers and clients, deployed production agents on AWS Bedrock AgentCore, and set up an evaluation framework on Langfuse that combines programmatic checks, LLM-as-judge scoring and human calibration. Newer ideas are tested first in the Engineering Lab on this site, where experiments are labelled as experiments.

I take on freelance and consulting work where I can own a meaningful piece of the outcome — an AI feature, a data pipeline, a backend, or an architecture review before a team commits to a direction.

Specializations

What I work on

  • Document intelligence

    Turning long, messy, inconsistent documents into structured data that downstream systems can trust — extraction, classification and validation.

  • Retrieval-augmented generation

    Chunking, embedding, hybrid retrieval, re-ranking and grounded generation, with evaluation sets that tell you when it gets worse.

  • Agentic systems

    Tool-calling agents, MCP integrations and multi-step workflows, designed with clear boundaries and a human in the loop where it matters.

  • Data platforms

    PySpark and Glue pipelines, Lakehouse migrations, standardization services and the quality gates that keep bad data out.

  • Python backends

    FastAPI services, integrations and authentication — the unglamorous layer that makes an AI feature a product.

  • Production delivery

    Containers, CI/CD, prompt versioning, cost tracking and observability, so the system keeps working after the demo.

Experience

Timeline

  1. Nov 2023 — Present

    Senior Software Engineer / Senior AI Engineer · Solugenix

    GenAI and agentic AI delivery for regulated enterprise clients: a benefits document intelligence platform for a global consulting client and a finance platform for a payments and data client, alongside AWS data engineering on a banking data migration.

    • Own end-to-end delivery of a benefits document intelligence platform, from architecture through production rollout, using Azure OpenAI, LangChain and FastAPI. It now processes 5,000+ policies a month.
    • Built a production RAG pipeline pairing Azure AI Search semantic retrieval with LLM-based extraction, cutting policy extraction time by 93% and manual review effort by 80%+.
    • Traced and fixed field hallucinations in benefits schedules by moving from single-query retrieval to multi-query retrieval combined with hybrid search.
    • Implemented Model Context Protocol (MCP) servers and clients to expose internal data sources and tools to LLM agents in a standard, reusable way.
    • Built and deployed production AI agents on AWS Bedrock AgentCore for the finance platform.
    • Designed an LLM evaluation framework from scratch on self-hosted Langfuse, using three scorer layers: programmatic evaluators, LLM-as-judge and human calibration.
    • Instituted LLMOps practices — prompt versioning, model evaluation, cost tracking, semantic caching and observability — improving throughput by 25%.
    • Core contributor on a SQL Server-to-PostgreSQL banking data migration on AWS Glue and PySpark, including a column-standardization engine (7 REST endpoints) that drives a config-as-data Glue job.
    • Led the document platform's data-layer migration to a Lakehouse architecture on Databricks.
    • Built REST APIs connecting the platform to 3+ enterprise systems; containerized with Docker and deployed to Kubernetes.
    Azure OpenAILangChainMCPAWS Bedrock AgentCoreLangfuseFastAPIPySparkAWS GlueDatabricksKubernetes

  2. Apr 2023 — Aug 2023

    Senior Software Developer · Pinnacle Data Reliability

    Built an analytics and visualization platform ingesting real-time data from 10+ sources, used daily by around 200 people.

    • Introduced Azure Data Services into the platform, speeding up data processing by roughly 60%.
    • Added GenAI-powered predictive analytics to the forecasting workflow.
    PythonAzure Data ServicesAnalyticsGenAI

  3. Dec 2020 — Apr 2023

    Technical Consultant · Netscore Technologies

    Data preparation and engineering for quantitative portfolio-management workloads.

    • Cleaned and prepared 2M+ records for Monte Carlo simulations used in investment decision-making.
    PythonData preparationSQL

  4. Apr 2019 — Feb 2020

    Software Engineer · Privus Technologies

    Built secure backend systems for business-critical operations.

    PythonBackend

  5. Mar 2017 — Apr 2019

    Early career — data quality & analytics · Amazon · Wipro (Google account)

    Validated and quality-checked large training datasets for ML model development at Amazon, and used SQL and Looker to detect and prevent fraud across millions of transactions at Wipro.

    SQLLookerML data quality

Philosophy

Engineering principles

Reliable products come from habits more than tools. These are the ones I hold to.

  1. 01

    Start from the outcome, not the model

    The first question is what decision or task gets faster or better. The model, framework and architecture follow from that.

  2. 02

    Measure before you trust

    LLM systems need an evaluation set the same way code needs tests. If accuracy cannot be measured, it cannot be improved or defended.

  3. 03

    Prefer the simplest design that works

    A well-built retrieval pipeline beats a clever agent that nobody can debug. Complexity is added only when the problem demands it.

  4. 04

    Data quality is the product

    Most AI failures are data failures. Validation, lineage and standardization come before the prompt.

  5. 05

    Design for the day after launch

    Logging, cost visibility, versioned prompts and clear failure modes are part of the build, not an afterthought.

  6. 06

    Say what is not known

    Clear scoping, honest estimates and early warnings about risk are worth more to a client than optimistic promises.

Have an AI or Data Engineering challenge? Let's turn it into a working product.

Taking on a limited number of AI and data engineering projects. Typical reply within 2 business days.

Let's Build SomethingDownload Resume